Here's a conversation I find myself having more and more:

A GTM leader — smart, well-resourced, committed to the initiative — walks me through everything their team is doing with AI. Typically, it’s a lot. They’ve adopted tools, teams and individuals have developed workflows, and they are quickly and efficiently creating excellent content. Which is all great.

Then I ask the question that tends to change the energy in the room: "What's moved on the revenue dashboard?"

Long pause.

"We're still early." Or: "The data isn't clean enough to attribute yet." Or, most honestly: "We're not sure."

This is the activity-performance gap. And it is the defining challenge of AI in GTM right now.

Activity is real. It's just not the point.

I want to be clear: AI activity is not nothing. Tools don't deploy themselves. Workflows don't get built without effort. Getting a team to consistently use new technology is hard, and the organizations doing it deserve credit.

But activity is an input, not an outcome. And somewhere along the way, in the rush to show AI progress, a lot of organizations started treating inputs as if they were outputs.

Seats activated. Content generated. Sequences launched. Pilots completed. These things go in the QBR deck, they satisfy the board's "what are you doing about AI" question, and they create the internal feeling of momentum.

What they don't do, on their own, is move the revenue metrics. And if the revenue metrics aren't moving, the activity is, at best, promising. At worst, it's expensive distraction.

Performance is a different animal entirely

AI performance shows up where it counts: pipeline velocity, conversion rates, time-to-close, rep productivity, cost per qualified opportunity, retention. The numbers that live on the revenue dashboard, not the AI initiative dashboard.

To be sure, it's slower to materialize. You're not going to see it in week two of a pilot. You won't be able to point to a single tool and say, "that's what did it." Performance is the cascading result of strategy, process, people, and technology all working together, and it takes time to show up.

But when it does show up, it's unambiguous. Deals close faster. Reps carry more pipeline. Conversion improves. The numbers that matter to the business change in ways that are visible, attributable, and sustainable.

That's the goal.

Why the gap exists — and it's not what most people think

The easy explanation for the activity-performance gap is that the tools aren't good enough yet, or that teams aren't using them correctly. Both of those things can be true. Neither of them is usually the real problem.

The reality is, the gap is almost always a design problem. Three specific design failures show up again and again:

The first is measuring the wrong things. When you can't easily measure performance impact, you measure what you can see — usage, activation, output volume. Those metrics are real, and they're useful early indicators. But when they become the definition of success rather than the leading indicators of it, the initiative drifts. Teams optimize for the metric they're being held to. If the metric is seats activated, you get seats activated. But, if the metric is pipeline velocity, you get something much more useful.

The second is deploying without designing. A tool gets rolled out – sometimes with good training, sometimes with a launch email and a link – but without a clear answer to the question that really matters: how exactly does this tool positively change how a rep works, and how does that change produce a better business outcome? If you can't draw that line before deployment, you're hoping the tool figures it out on its own. It won't.

The third is nobody owning the connection between AI and revenue. RevOps owns the tools. Sales owns the number. Marketing owns the pipeline. Leadership owns the initiative. And somehow, in the middle of all of that, nobody owns the question of whether the tools are actually moving the metrics. Accountability diffusion is not unique to AI, but it is particularly damaging here because performance requires someone to connect the disparate pieces of the puzzle.

The test worth running

Before your next AI initiative, or as an audit of the ones already running, try this: pick any AI tool or program your team is currently using, and draw a line from its output to a specific metric on your revenue dashboard. Connect activity to outcome.

How many steps does it take? If it's more than three, the initiative is probably too far upstream to drive measurable performance in the near term. That's not necessarily a reason to stop — some investments are the right ones even if the payoff is longer — but it's a reason to be honest about your timeline and not expect Q3 results from a Q1 deployment.

If you can't draw the line at all, that's the more important finding. It means the initiative wasn't designed with a performance outcome in mind. And the path forward isn't to deploy harder, but rather to step back and redesign the initiative around specific outcomes.

The question that reframes everything

The single question that I've found most useful in shifting a team from activity mode to performance mode is deceptively simple:

"What would have to be true about how this is used for it to show up in our revenue results in 90 days?"

That question forces specificity. It connects deployment to outcome. It creates a shared picture of what success actually looks like, not "we launched the pilot" but rather "our conversion rate from stage 2 to stage 3 improved by X points because reps are now doing Y differently."

It's a harder question than, "did the pilot succeed?" But it's the only one that leads somewhere worth going.

Activity is easy to generate. Performance is harder to earn. The organizations that figure out the difference, and build their AI programs around it, are the ones that look back in two years and can actually point to what changed.